A Deep Neural Architecture for Real-time Node Scheduling in Uplink Cell-Free Massive MIMO

This paper proposes a novel real-time and low-complexity solution for the joint power allocation and access point scheduling problem in the uplink cell-free massive multiple-input multiple-output (MIMO) with a serial bandwidth-limited fronthaul architecture. We devise a hybrid optimization framework based on a novel deep neural network architecture for access point scheduling and a low-complexity convex formulation for power allocation. The simulation results demonstrate the effectiveness of the proposed solution in which the trained network exhibits competitive performance compared to state-of-art optimization algorithms, however with a significant computation time reduction. It will be also discussed that the neural-based architecture is very advantageous for dynamic massive MIMO with time-varying number of users.